Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Beholder tab not showing up in TensorBoard app #55

Open
mathnathan opened this issue Nov 1, 2017 · 5 comments
Open

Beholder tab not showing up in TensorBoard app #55

mathnathan opened this issue Nov 1, 2017 · 5 comments

Comments

@mathnathan
Copy link

Since Beholder may not be 'officially' supported until TensorBoard version 0.5 as @jart suggested here, I am requesting help here for now.

I am attempting to use the latest TensorBoard build to access the beholder plugin. I clone TensorBoard and build it...

git clone https://github.com/tensorflow/tensorboard`
cd tensorboard
bazel build //tensorboard

I access Beholder as follows

FILE_LOC = os.path.dirname(os.path.realpath(__file__)) # Ensure use of
sys.path.insert(1,os.path.join(FILE_LOC, '../tensorboard/')) # local TensorBoard

from tensorboard.plugins.beholder.beholder import Beholder
beholder_writer = Beholder(session=sess, logdir=RUN_DIR)

and in my training loop I have

beholder_writer.update()

The code is executed and I see the beholder logs being written to my cpve_logs during training...

ls cpve_logs/plugins/beholder/
config.pkl  frame.summary  section-info.pkl

I execute the local tensorboard app built via bazel

./tensorboard/bazel-bin/tensorboard/tensorboard --logdir=cpve_logs

and I do not see the Beholder tab in the TensorBoard app.

I then attempted to use the TensorBoard version provided by @jart in tb-nightly.

pip install tb-nightly
tensorboard --logdir=cpve_logs

The TensorBoard app resulting from this build also does not yield the Beholder tab.

What am I doing wrong, and how can I access Beholder via TensorBoard?

@chrisranderson
Copy link
Owner

I don't know what has been happening to Beholder since it was merged - it may not be hooked into TensorBoard's nightly build at all yet. At the beginning, I was just handing the code over without integrating it into TensorBoard. I'm leaving that to the TensorBoard folks.

The only way I know how to use it is via this repository as of right now. Have you tried following the instructions in this repo?

@mathnathan
Copy link
Author

I have not. My initial goal was to access beholder in a future conscious way, such that it will be compatible with TensorBoard. It seemed to me that jart released the tb-nightly specifically with this in mind. I will give the instructions in this repo a try. Since the actual issue is not resolved, perhaps we can leave this open and see if someone else, perhaps @jart, has anything to say about it. Thanks for the hasty reply.

Cheers,

@chrisranderson
Copy link
Owner

Sadly, this repo might not be up to date, I'm not sure. I may have time Friday or Saturday to help you out if things don't work.

@mathnathan
Copy link
Author

Thank you for your diligence. I trust I can get it working, but I will report back should there be any further issues. Just wanted to be conscious of the merge with TensorBoard. I will wait until it is officially supported.

@jart
Copy link
Contributor

jart commented Nov 1, 2017

We'll be getting it working in the main repo over the next few months. It will probably be included in a release in Q1 2018, although we're planning on setting up a PIP package named tb-nightly so you might be able to use it sooner.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants